Information Processing Apparatus, Information Processing Method, and Program

1. An information processing apparatus comprising: a processor and associated computer memory to process a connection network storage means for storing a connection network which includes a first self-organization map and a second self-organization map each including a plurality of nodes and which also includes connection weights indicating connection strengths of nodes between the first self-organization map and the second self-organization map; first learning means for learning the first self-organization map, based on a first parameter extracted from an observed value output by observation means that observes an external world and outputs the observed value; winner node determination means for detecting a node having highest likelihood that the first parameter is observed at the node in the first self-organization map and determining the detected node as a winner node; searching means for searching the second self-organization map for a node having highest connection strength with the winner node and employing the detected node as a generation node; parameter generation means for generating a second parameter from the generation node; modification means for modifying the second parameter generated from the generation node; determination means for determining whether an end condition is satisfied to end modification of the second parameter, the modification being performed in accordance with the winner node determined for a value which is observed by the observation means when driving means performs a driving operation in accordance with the second parameter; first connection weight modification means for modifying the connection weight when the end condition is satisfied; second connection weight modification means for modifying the connection weight when evaluation by a user on the result of driving performed by the driving means is given as a reward by the user; and second learning means for learning the second self-organization map based on the second parameter obtained when the end condition is satisfied.

2. The information processing apparatus according to claim 1 , wherein the reward is a positive reward or a negative reward; and the second connection weight modification means performs the modification such that the connection weight is increased when the positive reward is given by the user, while the connection weight is decreased when the negative reward is given by the user.

3. The information processing apparatus according to claim 2 , wherein the second connection weight modification means modifies the connection weight such that the ratio of increasing or decreasing the connection weight by the second connection weight modification means is greater than the ratio of increasing or decreasing the connection weight by the first connection weight modification means.

4. The information processing apparatus according to claim 2 , wherein the second connection weight modification means modifies by the connection weight by changing the ratio of increasing or decreasing the connection weight, depending on the number of times the connection weight has been modified.

5. An information processing method comprising the steps of: based on a first parameter extracted from an observed value output by observation means that observes an external world and outputs the observed value, learning a first self-organization map stored in connection network storage means that stores a connection network which includes a first self-organization map and second self-organization map each including a plurality of nodes and which also includes connection weights indicating connection strengths of nodes between the first self-organization map and the second self-organization map; determining a winner node by detecting a node having highest likelihood that the first parameter is observed at the node in the first self-organization map and determining the detected node as the winner node; searching the second self-organization map for a node having highest connection strength with the winner node and employing the detected node as a generation node; generating a second parameter from the generation node; modifying the second parameter generated from the generation node; determining whether an end condition is satisfied to end modification of the second parameter, the modification being performed in accordance with the winner node determined for a value which is observed by the observation means when driving means performs a driving operation in accordance with the second parameter; modifying the connection weight when the end condition is satisfied; modifying the connection weight when evaluation by a user on the result of driving performed by the driving means is given as a reward by the user; and learning the second self-organization map based on the second parameter obtained when the end condition is satisfied.

6. A program to be executed by a computer embodied on a computer readable medium, the program comprising the steps of: based on a first parameter extracted from an observed value output by observation means that observes an external world and outputs the observed value, learning a first self-organization map stored in connection network storage means that stores a connection network which includes a first self-organization map and second self-organization map each including a plurality of nodes and which also includes connection weights indicating connection strengths of nodes between the first self-organization map and the second self-organization map; determining a winner node by detecting a node having highest likelihood that the first parameter is observed at the node in the first self-organization map and determining the detected node as the winner node; searching the second self-organization map for a node having highest connection strength with the winner node and employing the detected node as a generation node; generating a second parameter from the generation node; modifying the second parameter generated from the generation node; determining whether an end condition is satisfied to end modification of the second parameter, the modification being performed in accordance with the winner node determined for a value which is observed by the observation means when driving means performs a driving operation in accordance with the second parameter; modifying the connection weight when the end condition is satisfied; modifying the connection weight when evaluation by a user on the result of driving performed by the driving means is given as a reward by the user; and learning the second self-organization map based on the second parameter obtained when the end condition is satisfied.

7. An information processing apparatus comprising: a connection network storage unit that stores a connection network which includes a first self-organization map and a second self-organization map each including a plurality of nodes and which also includes connection weights indicating connection strengths of nodes between the first self-organization map and the second self-organization map; a first learning unit that learns the first self-organization map, based on a first parameter extracted from an observed value output by an observation unit that observes an external world and outputs the observed value; a winner node determination unit that detects a node having highest likelihood that the first parameter is observed at the node in the first self-organization map and determines the detected node as a winner node; a searching unit that searches the second self-organization map for a node having highest connection strength with the winner node and employs the detected node as a generation node; a parameter generation unit that generates a second parameter from the generation node; a modification unit that modifies the second parameter generated from the generation node; a determination unit that determines whether an end condition is satisfied to end modification of the second parameter, the modification being performed in accordance with the winner node determined for a value which is observed by the observation unit when a driving unit performs a driving operation in accordance with the second parameter; a first connection weight modification unit that modifies the connection weight when the end condition is satisfied; a second connection weight modification unit that modifies the connection weight when evaluation by a user on the result of driving performed by the driving unit is given as a reward by the user; and a second learning unit that learns the second self-organization map based on the second parameter obtained when the end condition is satisfied.